A Biparatopic Antibody That Modulates MET Trafficking Exhibits Enhanced Efficacy Compared with Parental Antibodies in MET-Driven Tumor Models

The MET receptor tyrosine kinase (RTK) gene is mutated and/or amplified in subsets of several human cancers (1). Recent clinical trials demonstrate that MET tyrosine kinase inhibitors (TKI) are active in these patient populations, validating MET as a cancer driver. In particular, lung cancers that harbor either MET exon 14 skipping mutations (MET-ex14) or high-level MET amplification are responsive to MET TKIs (2–7). Furthermore, recent findings suggest that MET amplification is an important driver of acquired resistance to EGFR TKIs in EGFR-mutated lung cancer (8–10).

Although TKIs can in some cases effectively block oncogenic RTKs, they are often nonselective, decreasing their tolerability. Thus, monoclonal antibodies have been explored as alternative therapeutics for RTK-driven cancers. Because many conventional MET antibodies that block binding of the MET ligand HGF are themselves activators of MET signaling, a monovalent MET antibody (onartuzumab) that lacks agonist activity was generated (11). However, when tested in phase III trials in lung and gastric cancers, onartuzumab failed to improve overall survival (12, 13). Given that MET appears to be a cancer driver primarily when mutated or amplified, the onartuzumab clinical failures likely resulted from the fact that these trials did not select for cancers that harbor MET genetic alterations. Furthermore, onartuzumab appears to be a poor blocker of ligand-independent MET signaling (14), which would limit its efficacy in MET-amplified cancers.

In an attempt to block MET function more effectively, several groups have generated antibodies (emibetuzumab, ABT-700, ARGX-111, SAIT301, and Sym015) that promote MET degradation, enabling efficacy in MET-amplified tumor models where MET signals, at least partially, in a ligand-independent fashion (15–20). Furthermore, several of these MET antibodies also show efficacy in preclinical models that harbor MET-ex14 mutations (15, 17, 20, 21). Although multiple MET antibodies have entered the clinic in recent years (21–27), it is unclear how effective these antibodies will be in MET-altered cancers.

In this article, we describe the generation of a biparatopic antibody in which each arm binds to a distinct epitope of MET. Our biparatopic antibody is highly effective in MET-dependent tumor models and functions at least partly through inhibition of MET recycling. Our results highlight the therapeutic relevance of modulating oncogenic receptor trafficking and, more generally, demonstrate that biparatopic antibodies can exhibit therapeutic properties not present in a mixture of the two parental antibodies.

Fully human antibodies against the MET extracellular domain were generated in VelocImmune mice using methods described previously (28, 29). Biparatopic MET antibodies were generated using methods described previously (30). In-house versions of emibetuzumab (comparator Ab1; patent US8,217,148B2), ABT-700 (comparator Ab2; patent US8,545,839B2), Sym015 (comparator Ab3; patent WO2016/042412A1), and onartuzumab (MET monovalent Ab; patent US7,892,550B2) were generated from their published primary sequences and were produced in CHO-K1 cells at Regeneron. Blocking antibodies against EGFR and ERBB3 were generated using VelocImmune mice and have been described previously (28).

Recombinant human HGF was obtained from R&D Systems. Cycloheximide was from Sigma-Aldrich Inc. Capmatinib was obtained from Selleck Chemicals.

All human cancer cell lines used in the study were obtained from ATCC and authenticated by short tandem repeat profiling (IDEXX Bioresearch). Cell growth in monolayer or in soft agar following various treatments was determined by measuring the reduction of the indicator dye alamarBlue (Invitrogen) according to the manufacturer's instructions.

Cell lysates (50 mmol/L Tris pH7.4, 150 mmol/L NaCl, 0.25 mmol/L EDTA, 1% Triton) were resolved on 4% to 20% Tris-Glycine gels and transferred to PVDF membranes (Novex). The following antibodies were obtained from Cell Signaling Technologies and used for primary labeling of Western blots: MET (D1C2), Phospho-MET (Y1234/1235), β-Tubulin, Phospho-p44/42 MAPK (T202/Y204), and p44/42 MAPK (Erk1/2). Secondary labeling of the immunoblots was performed with horseradish peroxidase-conjugated antibody followed by chemiluminescence detection (GE Healthcare).

MET antibodies were conjugated with Alexa-Fuor647 (AF647) according to the manufacturer's instructions. MET internalization was assessed by incubating EBC1 cells with AF647-labeled antibodies (10 μg/mL) at 4°C for 60 minutes. Unbound antibody was washed off, and endocytosis was initiated by transferring the cells to 37°C for various times. At each time point, antibodies remaining on the cell surface were detected with AF488-conjugated goat anti-human Fab fragment (4 μg/mL; Jackson ImmunoResearch Laboratories). Images were acquired with a Zeiss spinning disk confocal microscope, and the mean fluorescence intensity was measured on a pixel-by-pixel basis for all sections of the confocal stack. Surface MET at each time point is expressed as the percentage of the time zero level (100%) of the AF488/AF647 mean fluorescence intensity ratio.

To assess MET antibody recycling, AF647-labeled antibodies were bound to EBC1 cells at 4°C. Cells were then transferred to 37°C for 30 minutes to allow antibody internalization. Unbound antibody was washed off, and any labeled MET antibody remaining on the cell surface was blocked with unlabeled anti-human Fab (4 μg/mL). To label recycling MET antibodies, cells were incubated with AF488-labeled anti-human Fab (4 μg/mL) for 30 minutes at 4°C and then transferred to 37°C for 0 to 60 minutes. Confocal images were acquired, and colocalization was quantified on a pixel-by-pixel basis for all sections of the confocal stack. Antibody recycling at each time point is expressed as the AF488/AF647 mean fluorescence intensity ratio normalized to the time zero level.

The SEC-MALS system consists of an Agilent 1200 Series HPLC system equipped with an ultraviolet diode array detector coupled to a Wyatt Technology MiniDawn TREOS laser light scattering detector and an Optilab REX differential refractometer detector. Antibodies at 50 μmol/L were incubated with 50, 100, or 200 μmol/L MET ectodomain for 1 hour at ambient temperature. The samples were then injected into a pre-equilibrated Superose 6 Increase 10/300 GL size exclusion column at a volume equivalent to 100 μg of antibody at a flow rate of 0.5 mL/minute. The molar masses of the free antibodies, MET ectodomain, and antibody:MET ectodomain complexes were determined using Astra software (Wyatt Technology).

EM was performed using an FEI Tecnai T12 electron microscope, operating at 120 keV equipped with an FEI Eagle 4k × 4k CCD camera. Each sample was imaged over a layer of continuous carbon supported by nitrocellulose on a 400-mesh copper grid. Negative stain grids were transferred into the electron microscope using a room temperature stage. Images of each grid were acquired at multiple scales to assess the overall distribution of the specimen. High magnification images were acquired at nominal magnifications of 110,000× (0.10 nm/pixel) and 67,000× (0.16 nm/pixel). The images were acquired at a nominal underfocus of −2 μm to −1 μm and electron doses of ∼25–30e−/Å2.

Particles were identified in the high magnification images prior to alignment and classification. The individual particles were then selected, boxed out, and individual subimages were combined into a stack to be processed using reference-free classification. Individual particles in the 67,000× high magnification images were selected using automated picking protocols. An initial alignment was performed for each sample and class averages from that alignment that appeared to contain real particles were selected for additional rounds of alignment. Particle alignment and classification were conducted using a reference-free alignment strategy based on the XMIPP processing package. Algorithms in this package align the selected particles and sort them into self-similar groups or classes.

Antibodies at concentrations ranging from 5.2 pmol/L to 25 μmol/L were incubated with 0.125 nmol/L human MET ectodomain at room temperature and then transferred to ELISA plates coated with HGF. Plate-bound human MET (myc-tagged) was detected using horseradish peroxidase-conjugated c-myc secondary antibody and visualized using the colorimetric HRP substrate TMB, according to the manufacturer's recommendations.

Antibodies at concentrations ranging from 1.7 pmol/L to 100 nmol/L were added along with 300 pmol/L human HGF to triplicate wells of HEK293 cells engineered to express luciferase under the control of a serum response element and incubated at 37°C for 4 hours. ONE-Glo luciferase substrate was added to each well and relative light units were measured on a VICTOR X5 multilabel plate reader.

All mouse experiments were conducted in accordance with the guidelines of the Regeneron Institutional Animal Care and Use Committee. C.B.-17 SCID mice (6–8 weeks old) bearing established tumors were randomized such that the average tumor size and variance of the treatment groups were equivalent. Antibodies were administered by subcutaneous injection once every 3 to 4 days throughout the studies, and tumor volumes were monitored with caliper measurements.

We generated human MET antibodies with the goal of effectively blocking both ligand-dependent and ligand-independent MET signaling, thereby enabling efficacy in tumors that harbor MET genetic alterations. In initial screening, we identified antibodies that inhibited the growth of MET-amplified cancer cell lines, presumably by promoting MET degradation. However, these conventional antibodies generally exhibited significant agonist activity (Supplementary Table S1), likely reducing their utility as therapeutics. Therefore, in an effort to obtain an antibody with minimal agonist activity and the ability to promote MET degradation, we produced a series of biparatopic MET×MET antibodies in which each arm of the antibody recognizes a distinct epitope of MET. Because a single MET receptor can potentially be bound by two separate biparatopic antibody molecules, such antibodies can theoretically promote the formation of large antibody/receptor complexes that are efficiently internalized and degraded (31, 32). Furthermore, a biparatopic antibody might form a complex with MET that is not permissive of receptor activation, even if the parental antibodies are effective agonists.

One of our biparatopic MET antibodies inhibited the growth of EBC1 (MET-amplified NSCLC), Hs746T (MET-amplified and MET-ex14 gastric cancer), and SNU5 (MET-amplified gastric cancer) cells more potently than its parental antibodies (P Ab1, P Ab2), alone or in combination (Fig. 1A and B; Supplementary Fig. S1A). Importantly, the MET×MET antibody promoted apoptosis of EBC1 cells, indicating that the decrease in viable cell number does not simply reflect inhibition of proliferation (Supplementary Fig. S1B). The biparatopic antibody did not substantially reduce the growth of normal human bronchial epithelial cells or NCI-H358 NSCLC cells, which lack MET genetic alterations (Supplementary Fig. S1C–S1E).

To assess the activity of our biparatopic antibody relative to MET antibodies that have entered clinical trials, we used the publicly available sequences of emibetuzumab (comp Ab1), ABT-700 (comp Ab2), and Sym015 (comp Ab3, consisting of a mixture of two antibodies) to generate in-house versions of these antibodies. Our MET×MET antibody inhibited the growth of Hs746T cells more effectively than the three comparator antibodies (Fig. 1C). The potent activity of our biparatopic antibody prompted further investigation into its suitability as a therapeutic candidate. The kinetic binding parameters for the interactions of the biparatopic and parental antibodies with the MET ectodomain as well as confirmation that the two arms of the biparatopic antibody bind to nonoverlapping epitopes of MET are shown in Supplementary Figs. S2 and S3 and Supplementary Table S2, respectively.

Consistent with the increased ability of the biparatopic MET antibody to inhibit growth of MET-dependent cancer cells, the MET×MET antibody decreased MET protein levels to a greater extent than the parental antibodies alone or in combination (Fig. 1D; Supplementary Fig. S1F). Furthermore, our biparatopic antibody decreased MET protein levels more effectively than the three clinical-stage comparator antibodies in MET-ex14 Hs746T and NCI-H596 cells (Fig. 1E).

To assess the stability of MET in the presence of MET antibodies, total MET protein levels were measured in cells exposed to the protein synthesis inhibitor cycloheximide for various times. In MET-ex14 Hs746T cells, the biparatopic antibody, but not the parental antibodies, significantly reduced MET stability versus control antibody (Fig. 1F). In EBC1 cells, the parental antibodies and the biparatopic antibody decreased MET stability versus control antibody, with the biparatopic antibody having a more pronounced effect (Supplementary Fig. S4).

The ability of the biparatopic antibody to promote MET degradation more effectively than its parental antibodies suggests that the biparatopic antibody traffics efficiently to lysosomes. To test this hypothesis, fluorescently-labeled MET antibodies were used to measure internalization rates in EBC1 cells. Both the biparatopic antibody and parental antibody 2 were rapidly internalized, with ∼50% of the antibody/MET complexes internalized within 20 minutes (Fig. 2A). Parental antibody 1 was internalized more slowly, while a monovalent MET antibody (an in-house version of onartuzumab) largely remained on the cell surface.

Following internalization to early endosomes, RTKs either recycle back to the cell surface or traffic through late endosomes to lysosomes (33). To evaluate the recycling of internalized antibody/MET complexes, the fraction of fluorescently labeled antibody returning to the plasma membrane was measured over time (upon return to the cell surface, the antibody/MET complex is bound by a second labeled antibody present in the culture medium). The biparatopic MET antibody recycled back to the cell surface less efficiently than either parental antibody, with a 25% increase in the signal of recycled biparatopic antibody over 60 minutes, compared with a 40% to 60% increase for the parental antibodies (Fig. 2B). Thus, although both the biparatopic antibody and parental antibody 2 are rapidly internalized, recycling of the biparatopic antibody is significantly less efficient, likely explaining its enhanced ability to promote MET degradation.

To assess the kinetics of antibody trafficking to lysosomes, the degree of endosome (transferrin-labeled) and lysosome (dextran-labeled) occupancy by fluorescently labeled MET antibodies was measured shortly after antibody addition. Confocal imaging demonstrated that the biparatopic MET antibody rapidly traffics to both endosomes and lysosomes (by 15 minutes), while parental antibody 1 preferentially colocalized with the endosomal marker at this early time point (Supplementary Fig. S5). Parental antibody 2 exhibited an intermediate phenotype. Together, our data suggest that the biparatopic antibody promotes more effective degradation of MET than the parental antibodies by preferentially trafficking to lysosomes rather than recycling back to the cell surface.

One possible explanation for the ability of biparatopic antibodies to promote effective target degradation is that they promote formation of large antibody/receptor complexes (greater than the expected 1:2 complex for conventional antibodies). Therefore, we used multiple-angle light scattering coupled to size exclusion chromatography to determine the stoichiometry of the complexes that our antibodies form with the MET ectodomain. Although the chromatograms for parental antibodies 1 and 2 suggest that they form 1:2 and 1:1 antibody/MET complexes, respectively, the MET×MET antibody appeared to form a 2:2 complex (587.7 kDa; Fig. 3A).

We next used negative stain EM with two-dimensional class averaging to investigate the conformation of the biparatopic antibody/MET ectodomain complex. We observed circular particles of approximately 30 nm that appear to consist of two biparatopic antibodies bound to two MET ectodomains (Fig. 3B). The antibodies were oriented facing each other with opposing Fabs bound to the same MET N-terminal domain. The Fc portion of the antibodies as well as the IPT1-4 domains of MET are faint in appearance, likely due to high structural variability in these regions as a result of conformational flexibility (Fig. 3B). This mode of binding is consistent with hydrogen–deuterium exchange epitope mapping data showing that the biparatopic MET antibody recognizes two nonoverlapping epitopes on the Sema domain (Supplementary Fig. S6). A model for the 2:2 complex that incorporates the two-dimensional EM class average data and the epitope mapping data is depicted in Fig. 3C. Although the biparatopic MET antibody does not appear to form very large complexes with MET (at least in vitro), the observed 2:2 complex is larger than the complexes formed by the parental antibodies. It therefore remains possible that the increased size of the biparatopic antibody/MET complex contributes to its differential trafficking properties.

We next assessed whether the biparatopic MET antibody can rapidly inhibit ligand-independent MET signaling (in addition to promoting MET degradation over time). Baseline MET activation in EBC1 cells is ligand-independent, as shown by the inability of the monovalent MET antibody, a potent blocker of HGF binding (11), to inhibit MET phosphorylation (Fig. 4A). Although treatment of EBC1 cells with HGF resulted in modest increases in MET and ERK phosphorylation (ERK is a key mediator of MET-dependent biological responses; ref. 1), the MET×MET antibody resulted in rapid (by 30 minutes) inhibition of ERK phosphorylation (Fig. 4A). The decrease in ERK phosphorylation occurred without a decrease in total MET levels (which occurs more slowly) and, strikingly, without inhibition of MET phosphorylation. In contrast to the biparatopic antibody, administration of the parental antibodies alone or in combination for up to 60 minutes did not inhibit ERK phosphorylation.

As shown in Fig. 4B, the mechanism through which the biparatopic antibody inhibits downstream MET signaling is clearly distinct from that used by a MET TKI. The biparatopic antibody decreased ERK phosphorylation beginning at 15 minutes after initiation of treatment, with a further decrease apparent at 30 minutes. Again, this decrease in ERK phosphorylation occurred without any change in MET phosphorylation or total MET levels (Fig. 4B). In contrast, the selective MET TKI capmatinib substantially decreased MET phosphorylation by 10 minutes, accompanied by almost complete inhibition of ERK phosphorylation. Finally, in MET-amplified and MET-ex14 Hs746T cells, which exhibit constitutive ligand-independent MET activation, the biparatopic antibody inhibited ERK activation at 30 and 60 minutes after initiation of treatment without any change in MET phosphorylation or in total MET levels (Fig. 4C). Neither of the parental antibodies had as strong an inhibitory effect on ERK phosphorylation as the MET×MET antibody.

Thus, our biparatopic antibody rapidly inhibits downstream MET signaling without interfering with ligand-independent MET phosphorylation. One possible explanation for this observation is that the biparatopic antibody promotes MET trafficking to multivesicular endosomes, where RTKs are incorporated into intraluminal vesicles and physically separated from MAP kinase pathway components (see Fig. 4D for a model). As shown in Supplementary Fig. S3, the MET×MET antibody traffics rapidly to lysosomes in EBC1 cells (within 15 minutes), so the early blockade of ERK activation observed here (15–30 minutes) could plausibly reflect antibody-induced trafficking of MET to multivesicular endosomes.

To determine whether the biparatopic antibody can inhibit ligand-dependent MET activation, we first tested its ability to block HGF binding to MET in an ELISA-based assay. As shown in Fig. 5A, incubation of the biparatopic antibody with recombinant MET prevented binding to immobilized HGF with an IC₅₀ of 60 nmol/L (9 μg/mL). At the highest antibody concentration tested, the biparatopic antibody blocked 96% of MET binding to HGF, while parental antibodies 1 and 2 achieved 91% and 66% blockade, respectively. We next tested the effect of the antibodies on HGF-induced gene expression in a serum response element–driven luciferase reporter assay. HGF-induced luciferase activity was completely blocked by the addition of biparatopic MET antibody (IC₅₀ 0.8 nmol/L or 120 ng/mL) or parental antibody 2 (IC₅₀ 2.3 nmol/L or 345 ng/mL), whereas parental antibody 1 had no effect (Fig. 5B). (Although parental antibody 1 effectively prevents HGF binding, it activates MET signaling on its own.)

To assess the agonist properties of the biparatopic and parental antibodies, we used the HGF-responsive lung cancer cell line NCI-H596, which harbors a MET-ex14 mutation (34) but has no detectable MET activation at baseline due to a lack of HGF production. The addition of HGF to these cells induced strong MET phosphorylation as well as robust, sustained phosphorylation of ERK (Fig. 5C). In contrast, the biparatopic antibody and both parental antibodies induced a very modest, but detectable, increase in MET phosphorylation. Interestingly, the induction of ERK phosphorylation by the biparatopic antibody was transient, returning to near-baseline level within 1 hour of stimulation, whereas the ERK phosphorylation induced by parental antibodies 1 and 2 was more sustained, remaining well above baseline level for up to 2 hours. The substantial difference in the duration of the ERK signals induced by the biparatopic antibody versus the parental antibodies did not reflect differences in the levels of MET phosphorylation over time, which were equivalent (Fig. 5C). This rapid termination of ERK phosphorylation in NCI-H596 cells in response to the biparatopic antibody is reminiscent of the rapid blockade of ERK observed in EBC1 and Hs746T cells treated with the biparatopic antibody (Fig. 4) and could reflect inhibition of MET recycling and rapid trafficking to multivesicular endosomes. The very transient activation of downstream MET signaling in response to the biparatopic antibody could have important functional consequences, as the duration of ERK activation has been shown previously to be an important determinant of biological responses downstream of RTKs (35).

Consistent with the inability of the biparatopic antibody to drive sustained ERK activation, it failed to stimulate proliferation of NCI-H596 cells in soft agar, whereas both HGF and the parental MET antibodies (alone or in combination) promoted cell growth to varying degrees (Fig. 5D). We also assessed the ability of MET antibodies or HGF to promote the growth/survival of Cal27 cancer cells in the context of combined EGFR and ERBB3 blockade (we have previously shown that Cal27 cells are dependent on EGFR/ERBB3 signaling; ref. 28). Although both HGF and the parental MET antibodies were able to partially rescue the growth/survival of Cal27 cells, the MET×MET antibody had no significant effect (Fig. 5E).

We next tested the ability of the biparatopic antibody to inhibit tumor growth in a model (U-87MG glioblastoma) that is driven by autocrine HGF signaling (36). As shown in Fig. 5F, the MET×MET antibody at 10 mg/kg was as effective as the HGF-blocking monovalent MET antibody, with both agents promoting tumor regression. Thus, our biparatopic MET antibody potently blocks HGF-dependent MET signaling and exhibits no signs of biologically meaningful MET agonist activity, which would limit its antitumor effect in this HGF-driven model.

We next examined the antitumor efficacy of the biparatopic antibody in human tumors harboring MET genetic alterations. The MET×MET antibody at a dose of 1 mg/kg induced substantial regression of established MET-amplified SNU5 gastric cancer xenografts and induced complete and durable tumor regression at a dose of 10 mg/kg (Fig. 6A). In contrast, treatment with the monovalent MET antibody at 10 mg/kg resulted only in a modest and transient tumor regression, presumably reflecting ongoing ligand-independent MET signaling that is not effectively blocked by this antibody. Similarly, in the MET-amplified EBC1 NSCLC model, treatment with MET×MET antibody at 25 mg/kg completely suppressed tumor growth whereas the monovalent MET antibody had a very modest effect, even at this high dose (Fig. 6B). Furthermore, a comparator MET antibody (Comp Ab1) was less effective than the biparatopic antibody, only providing tumor growth delay.

In MET-amplified and MET-ex14 Hs746T xenografts, the biparatopic antibody at doses of 5 mg/kg and higher promoted complete tumor regression (Fig. 6C). The Hs746T tumors did regrow, however, with distinct kinetics depending on the dose, suggesting slightly different degrees of initial tumor cell killing. Importantly, as shown in Fig. 6D, our biparatopic antibody was substantially more effective than two clinical-stage MET antibodies (Comp Ab1, Comp Ab2) in this model.

Because the biparatopic antibody is more effective than its parental antibodies in vitro, we hypothesized that it would also exhibit greater antitumor efficacy. As shown in Fig. 6E, treatment with the biparatopic antibody at 5 mg/kg resulted in regression of Hs746T xenografts, whereas administration of the parental antibodies alone or in equimolar combination only resulted in tumor growth delay. Furthermore, in the SNU5 model, only the biparatopic antibody, but not the parental antibodies alone or in combination, was able to promote complete tumor regression (Fig. 6F). Thus, our biparatopic MET antibody has antitumor properties that are clearly distinguishable from those of the parental antibodies and the antibody mixture, making it a potentially attractive therapy for MET-driven cancers.

In this report, we have characterized the properties of a biparatopic MET antibody that is highly effective in MET-dependent tumor models, outperforming both monovalent and conventional MET antibodies that have previously entered the clinic. Importantly, even in the context of the MET trafficking defect that is thought to underlie the oncogenic potential of the exon 14 skipping mutations, the biparatopic antibody effectively promotes MET degradation and inhibits MET function. Our findings highlight the therapeutic potential of antibody-driven modulation of MET trafficking and, more generally, show that biparatopic antibodies can exhibit therapeutic properties distinct from those of a mixture of the two parental antibodies.

The concept of enhancing the internalization and degradation of receptors by promoting the formation of antibody/receptor clusters on the cell surface has been explored fairly extensively (19, 20, 31, 32, 37–43). The ability of antibodies to promote receptor degradation could result from an increased rate of receptor internalization, a decrease in receptor recycling or both. Our biparatopic antibody promotes MET internalization and inhibits recycling, but it is the effect on recycling that appears to distinguish the biparatopic antibody from the parental antibodies (Fig. 2). The notion that recycling of receptors can be inhibited by clustering is not a new one, as both antibody mixtures (42) and biparatopic DARPins (44) have been shown to inhibit RTK recycling. The simplest explanation for the ability of antibodies to divert RTKs from the recycling pathway is that antibody/receptor clusters are restricted from entering the small-diameter tubules that bud from sorting endosomes and carry cargo back to the cell surface (45, 46). This simple model is consistent with the enhanced ability of our biparatopic antibody (which forms a larger antibody/MET complex than the parental antibodies) to inhibit recycling. Decreased entry of biparatopic antibody/MET clusters into recycling tubules would lead to enhanced trafficking down the degradative pathway. However, we cannot rule out an effect of the MET biparatopic antibody on specific sorting mechanisms that regulate MET recycling (47) or incorporation into intraluminal vesicles (33).

Another key property of our biparatopic antibody is its ability to rapidly inhibit the constitutive ligand-independent MET/ERK signaling in MET-amplified cells (without inhibiting MET phosphorylation) and to induce very transient downstream ERK signaling in cells that exhibit no MET activation at baseline (Figs. 4 and 5). The inability of the biparatopic antibody to promote sustained downstream signaling is consistent with its failure to drive MET-dependent biological responses. We believe that this rapid termination of MET/ERK signaling may reflect incorporation of MET into the intraluminal vesicles of multivesicular endosomes, an important step in RTK signal termination (33, 48). The possibility that the inhibition of MET recycling by the biparatopic antibody is linked to inhibition of signaling is consistent with the general concept that MET trafficking modulates signaling output (49, 50) and the specific observation that MET recycling is required for sustained ERK activation (47).

Given recent positive clinical data with MET-directed TKIs, particularly in lung cancers harboring MET exon 14 mutations, MET is once again an attractive target, as evidenced by the number of clinical trials that are currently investigating MET targeted therapies. The ability of our biparatopic antibody to potently inhibit both ligand-dependent and ligand-independent MET signaling and its superior efficacy in MET-driven tumor models compared with other clinical-stage MET antibodies make it an attractive therapeutic candidate. This antibody is now in a phase I trial in MET-altered advanced NSCLC (NCT04077099).

J.O. DaSilva, K. Yang, A.E. Perez Bay, M.C. Franklin, D. Dudgeon, F.J. Delfino, T.B. Potocky, G. Chen, W.C. Olson, and G. Thurston are employees/paid consultants for Regeneron Pharmaceuticals Inc. No potential conflicts of interest were disclosed by the other authors.

Conception and design: J.O. DaSilva, A.E. Perez Bay, J. Andreev, A. Rafique, A. Dore, R. Babb, W.C. Olson, G. Thurston, C. Daly

Development of methodology: A.E. Perez Bay, D. Dudgeon, A. Rafique, A. Dore, F.J. Delfino, T.B. Potocky, G. Chen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Yang, A.E. Perez Bay, P. Ngoi, M.C. Franklin, D. Dudgeon, A. Dore, G. Chen

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.O. DaSilva, A.E. Perez Bay, P. Ngoi, M.C. Franklin, D. Dudgeon, A. Rafique, A. Dore, F.J. Delfino, T.B. Potocky

Writing, review, and/or revision of the manuscript: J.O. DaSilva, J. Andreev, E. Pyles, M.C. Franklin, A. Rafique, A. Dore, T.B. Potocky, G. Thurston, C. Daly

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): G. Chen, D. MacDonald

Study supervision: J.O. DaSilva, T.B. Potocky, W.C. Olson, C. Daly

This study was fully funded by Regeneron Phamaceuticals, Inc.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.